Bodvarsson and Gibson (1999) also argued that the rating scales used in existing tests of the
tipping–service relationship were ordinal and, therefore, provided data unsuitable for the statistical
tests performed. This argument has no merit. Numerical rating scales such as those used in evaluations
of service quality are commonly regarded as having equal (or nearly equal) intervals between response
options (Churchill, 1996). This assumption is clearly justified when only the end points of the numerical
response options are given verbal labels, because the numbered intermediate points on the scale
represent equal intervals between the two endpoints. Contrary to the impression that Bodvarsson and
Gibson (1999) gave, the rating scales used by Lynn and Grassman (1990) and others did not have verbal
labels attached to intermediate points on the scales. Thus, these numeric scales were legitimately
treated as interval level data. Even if the numeric scales were only ordinal as Bodvarsson and Gibson
suggest, they would still be acceptable for use in correlational analyses because such analyses are robust
with respect to monotonic transformations of data (Nunnally, 1978).